Camera calibration: adjusting camera lense distortion¶
by Uki D. Lucas
Motivation¶
The raw camera data usually has a certain level of distortion caused by lense shape, this is especially pronounced on the edges of the image. The correction is essential in applications like image recognition used in autonomous vehicles, robotics and even in 3D printing.
Solution approach¶
The common solution is to compare a known shape object e.g. a chessboard with the image taken, then calculate this specific camera's adjustment parameters that then can be applied to every frame taken by the camera. If the camera changes, the parameters have to be recalibrated.
In [1]:
#!/usr/bin/python3
import numpy as np
import cv2
import glob
import json
import matplotlib.pyplot as plt
%matplotlib inline
The pattern will look for "inner corners" e.g. black touching the black square¶
In [2]:
nx = 9 # horizontal
ny = 6 # vertical
In [3]:
# read a list of files using a parern
images = glob.glob("camera_cal/calibration*.jpg") # e.g. calibration19.jpg
print("found", len(images), "images" )
In [4]:
# Initialise arrays
# Object Points: 3d point in real world space
object_point_list = []
#Image Points: 2d points in image plane.
image_points_list = []
In [5]:
# Generate 3D object points
object_points = np.zeros((nx*ny, 3), np.float32)
object_points[:,:2] = np.mgrid[0:nx, 0:ny].T.reshape(-1, 2)
print("first 5 elements:\n", object_points[0:5])
In [18]:
# see: http://docs.opencv.org/trunk/dc/dbb/tutorial_py_calibration.html
termination_criteria = (cv2.TERM_CRITERIA_EPS + cv2.TERM_CRITERIA_MAX_ITER, 30, 0.001)
chessboard_dimentions = (nx, ny)
for file_name in images:
image_original = cv2.imread(file_name)
# convert BGR image to gray-scale
image_gray = cv2.cvtColor(image_original, cv2.COLOR_BGR2GRAY)
# Find the chess board corners
# Paramters:
# - image_gray
# - the chessboard to be used is 9x6
# - flags = None
has_found, corners = cv2.findChessboardCorners(image_gray, chessboard_dimentions, None)
print("processing:", file_name)
if has_found == True:
# fill in ObjectPoints
object_point_list.append(object_points)
corners2 = cv2.cornerSubPix(image_gray, corners, (11,11), (-1,-1), termination_criteria)
# fill in ImagePoints
image_points_list.append(corners2)
# Draw and display the corners
# I have to clone/copy the image because cv2.drawChessboardCorners changes the content
image_corners = cv2.drawChessboardCorners(image_original.copy(), chessboard_dimentions, corners2, has_found)
plt.figure()
plot_image = np.concatenate((image_original, image_corners), axis=1)
plt.imshow(plot_image)
plt.show()
else:
print("The", chessboard_dimentions, "chessboard pattern was not found in file", file_name)
plt.figure()
plt.imshow(image_original)
plt.show()
Calibrate using points¶
In [9]:
# See below to understand returned values
ret, matrix, distortion, rvecs, tvecs = cv2.calibrateCamera(
object_point_list,
image_points_list,
image_gray.shape[::-1],
None,
None)
result=dict()
result['ret']=ret
result['matrix']=np.array(matrix).tolist() # camera matrix
result['dist']=np.array(distortion).tolist() # distortion coefficients
result['rvecs']=np.array(rvecs).tolist() # rotation vectors
result['tvecs']=np.array(tvecs).tolist() # translation vectors
In [10]:
with open('calibrate_camera_output.json', 'w') as f:
json.dump(result, f, indent=4, sort_keys = True)
In [12]:
image_dimentions = image_gray.shape[:2] # height, width
new_matrix, roi = cv2.getOptimalNewCameraMatrix(matrix, distortion, image_dimentions, 1, image_dimentions)
Remove the distortion form the images¶
In [22]:
for file_name in images:
print("processing:", file_name)
image_original = cv2.imread(file_name)
image = cv2.undistort(image_original, matrix, distortion, None, new_matrix)
plt.figure(figsize=(20,10))
#plt.figure()
plot_image = np.concatenate((image_original, image), axis=1)
plt.imshow(plot_image)
plt.show()
# save to disk
if "calibration2.jpg" in file_name:
cv2.imwrite('image_undistorted_2.png', plot_image)
In [ ]: